CN111639526A - Power transmission line target detection and monitoring method based on deep learning - Google Patents
Power transmission line target detection and monitoring method based on deep learning Download PDFInfo
- Publication number
- CN111639526A CN111639526A CN202010323443.XA CN202010323443A CN111639526A CN 111639526 A CN111639526 A CN 111639526A CN 202010323443 A CN202010323443 A CN 202010323443A CN 111639526 A CN111639526 A CN 111639526A
- Authority
- CN
- China
- Prior art keywords
- model
- power transmission
- transmission line
- training
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 title claims abstract description 24
- 230000005540 biological transmission Effects 0.000 title claims abstract description 20
- 238000012544 monitoring process Methods 0.000 title claims abstract description 19
- 238000013135 deep learning Methods 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000011156 evaluation Methods 0.000 claims abstract description 9
- 230000000694 effects Effects 0.000 claims abstract description 8
- 230000008859 change Effects 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims abstract description 6
- 230000009467 reduction Effects 0.000 claims abstract description 5
- 238000012795 verification Methods 0.000 claims abstract description 5
- 238000010586 diagram Methods 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 6
- 230000006378 damage Effects 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 230000006872 improvement Effects 0.000 claims description 3
- 230000003321 amplification Effects 0.000 abstract description 2
- 238000003199 nucleic acid amplification method Methods 0.000 abstract description 2
- 238000005070 sampling Methods 0.000 description 4
- 238000004140 cleaning Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 239000012855 volatile organic compound Substances 0.000 description 2
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a power transmission line target detection and monitoring method based on deep learning, which is characterized in that after a large amount of image data are collected in a power transmission line target detection scene, the image data are cleaned and screened according to a certain positive and negative sample proportion, and then dangerous targets are accurately labeled manually and are divided into a training set, a verification set and a testing set; performing data amplification methods such as zooming, rotating, cutting, turning and the like before model training; constructing a model of the improved Huge-YOLO v3 on a Darknet deep learning framework, training by means of the latest powerful software and hardware computing equipment, storing the model according to loss reduction change in the training process, judging through some evaluation indexes, and finally, summarizing the index evaluation score and selecting the model with the best convergence fitting effect; the method can acquire higher-level semantic information and lower-level fine-grained characteristic information, thereby improving the target detection accuracy and solving the problem of huge difference of target scales.
Description
Technical Field
The invention relates to the technical field of power system monitoring, in particular to a power transmission line target detection and monitoring method based on deep learning.
Background
A large amount of image data are collected in a monitoring system for preventing external damage of a power transmission line, and a target with potential threat to the line can be detected and early warned through a deep learning target detection technology, for example, targets such as a person who is lifted by a steam, a fishing person and the like can cause interference damage to a wire. The deep learning model has strong representation capability and can play an important role in the technical field of image target monitoring. The object detection focuses on the category information and the position information of a specific object, and currently, a mainstream framework is mainly divided into two detection models, namely a two-stage (two-stage) detection model and a single-stage (one-stage) detection model.
YOLO v3 is based on a better basic classification network (class ResNet) and a classifier (Darknet-53), and simultaneously absorbs the multi-scale prediction idea of a pyramid feature representation method (FPN), YOLO v3 performs size transformation in the image tensor forward process, increases step size processing through a convolution kernel, scales for 3 times to reach 1/32, extracts three layers of feature maps (the sizes are respectively 52 × 52, 26 × 26 and 13 × 13) from different scales, performs independent prediction on each layer, performs splicing and fusion for further prediction in an up-sampling mode, finally performs prediction by using 9-scale anchor box clustered by dimensions, and performs prediction by evenly distributing the anchor box on 3 layers of feature maps, wherein the receptive fields of the anchor box correspond to 8 × 8, 16 × 16 and 32 × 32 respectively. The conversion of the YoLO v3 class loss function from Softmax to a Sigmoid function solves the same type of large and small classification problems. While YOLO v3 performed well in terms of speed, multi-scale training was not particularly effective in detecting object targets that were extremely small and large in order to strike a tradeoff between speed and accuracy.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a power transmission line target detection and monitoring method based on deep learning, which is improved on the basis of single-stage YOLO v3 and solves the problem that the shot images of a power transmission line scene contain large difference of target scale; the method for extracting and fusing the features at different layers of the YOLO v3 by multi-scale prediction is improved, and higher-layer semantic information and lower-layer fine-grained feature information can be obtained, so that the target detection accuracy is improved, and the problem of huge difference of target scales is solved.
The invention relates to a power transmission line target detection and monitoring method based on deep learning, which comprises the following steps:
(1) acquiring an image of a specific target through monitoring shooting equipment on a power transmission line site;
(2) screening the images in the step (1) according to a positive sample-negative sample ratio of 3: 7-5: 5, selecting images with proper angles, and deleting images without targets or with unclear angles, wherein the positive samples are images containing specific targets, and the negative samples are images without specific targets;
(3) labeling specific targets in the screened images by using a LabelImg tool, and dividing the images into a training set, a verification set and a test set;
(4) amplifying an image sample before model training, and amplifying the image by adopting a scaling, rotating, cutting or turning mode;
(5) constructing a model on a Darknet deep learning framework by using an improved Huge-YOLO v 3;
(6) training by means of a GPU Nvidia GTX 1080Ti hardware environment based on the constructed model; continuously adjusting parameters in the training process, optimizing the loss function loss value to the minimum, storing different models according to loss reduction change, performing comprehensive judgment through accuracy, recall rate, IoU and mAP evaluation indexes, and selecting the model with the best convergence fitting effect as a detection and monitoring model;
(7) inputting the image to be detected into a trained detection and monitoring model, carrying out model reasoning and identification on the type and position of the specific target of the image and the credibility of the target, sending early warning information to related power transmission line inspection workers through an alarm system when the specific target with potential harm appears, and processing the specific target according to the actual condition by the inspection workers.
The improvement of the Huge-Yolo v3 is that the size of an input image is changed from 416 × 416 × 3 to 832 × 832 × 3, a network layer is added on a main structure Darknet-53, and four feature maps of 104 × 104, 52 × 52, 26 × 26 and 13 × 13 are respectively extracted; and fusing the four-layer characteristic diagram by adopting a self-adaptive characteristic fusion mode to obtain fused characteristics, wherein the formula is as follows:
where y is the fusion feature, l is the resolution feature layer, ij represents the feature vector at position (i, j) on the feature map, and a, β, γ, λ are the spatial weight parameters.
The method has the advantages and the technical effects that:
the method of the invention can well adapt to the problem of large scale change based on the improved Huge-YOLO v3 model of YOLO v3, and the recognition accuracy effect is obviously improved; the semantic information of a higher layer and the fine-grained characteristic information of a lower layer can be obtained, so that the target detection accuracy is improved, and the problem that the target scale is greatly different is solved.
Drawings
FIG. 1 is a flowchart of Huge-YoLO v3 model training;
FIG. 2 is a diagram of a four-level feature map fusion process;
FIG. 3 is a diagram illustrating a specific object detected in an item;
FIG. 4 is a diagram illustrating a specific object detected in an item;
FIG. 5 is a diagram illustrating a specific object detected in a project.
Detailed Description
The present invention is further illustrated by the following examples, but the scope of the invention is not limited to the above-described examples. Example 1: as shown in fig. 1, the method for detecting and monitoring the target of the power transmission line based on deep learning comprises the following steps:
in this embodiment, a GPU Nvidia GTX 1080Ti hardware environment is used to complete a training test; adopting a Darknet compiling environment, wherein a Ubuntu system is required to be installed in a hardware environment, and meanwhile, relevant software (drivers, CUDA and CUDNN) such as NVIDIA and the like are installed for a GPU acceleration training model and an image processing relevant library such as OpenCV and the like;
to increase the size of large objects (e.g., a car crane), a larger receptive field and higher-level semantic information are needed, while to far away small objects (e.g., a person), fine-grained information in the underlying features is needed to be resolved. In the training process of the deep learning target detection model, the number of samples of data and the refinement degree of labels are one of three key factors for determining the fitting of the model. Therefore, after a large amount of image data are collected according to a power transmission line target detection scene, the image data are cleaned and screened according to a certain positive and negative sample proportion, and then dangerous targets are accurately marked manually and are divided into a training set, a verification set and a test set. The model training is preceded by data augmentation methods such as scaling, rotation, cropping, flipping, etc. Constructing a model of the improved Huge-YoLO v3 on a Darknet deep learning framework, training the model by means of the latest strong software and hardware computing equipment, storing the model according to loss reduction change in the training process, and judging according to some evaluation indexes, such as: accuracy (Precision), Recall (Recall), iou (interaction over union), mep (mean Average Precision), and the like. Finally, selecting a model with the best convergence fitting effect in conclusion of the index evaluation score;
1. data acquisition, namely uploading and collecting a large amount of image data through monitoring shooting equipment on the site of the power transmission line;
2. cleaning and screening data, namely cleaning and screening according to a positive sample and negative sample ratio of 5:5, specifically operating and selecting images with proper angles, deleting a large number of images without targets or with unclear and high repeatability, and ensuring data diversity, wherein the positive sample is an image containing a specific target;
3. data labeling, namely manually and accurately labeling a specific target in an image by using a LabelImg tool to generate a label file, wherein the data is in a VOC (volatile organic compound) format and is divided into a training set, a verification set and a test set;
4. and (3) data amplification, namely amplifying the data sample before model training, such as: the method comprises the following steps of scaling, rotating, cutting, turning and the like, and aims to increase the diversity of training data, improve the accuracy and generalization capability of a model and avoid overfitting;
5. model training, downloading pretraining weight based on ImageNet, constructing a model on a Darknet deep learning framework by using improved Huge-YoLO v3, and modifying a training configuration file, wherein parameters such as batch-64 and learning-rate-0.001 are used;
in order to adapt to improvement of mobile and embedded edge equipment based on YOLO v3, the Tiny-YOLO v3 greatly simplifies a network structure, changes a characteristic diagram from three layers into two layers, and ensures the speed in an environment with limited resources at the cost of reducing accuracy; and the speed requirement of the power transmission line target detection scene is not real-time detection, so that the accuracy can be greatly improved by depending on strong rear-end computational resources, and Huge-YOLO v3 is provided based on YOLO v 3.
Improved relation table of corresponding relation among feature diagram, receptive field and anchor box size
The method mainly comprises the steps of changing the size of an input image from 416 x 3 to 832 x 3, adding a network layer on a backbone structure Darknet-53, respectively extracting four Feature maps of 104 x 104, 52 x 52, 26 x 26 and 13 x 13, and adopting an adaptive Feature Fusion mode (ASFF) to Adaptively adjust and learn a Fusion weight through scale map resize to respectively fuse different Feature layers to a final Feature map as the size relations of the Feature maps, the receptive field and the anchor box in the table correspond to each other so as to more fully acquire high-level semantic information and bottom-level fine-grained information and not to follow the original up-sampling splicing mode;
as shown in FIG. 2, the dotted box represents the fusion process for the four-layer feature map, L1,L2,L3,L4Respectively corresponding to the characteristics of four layers of Level1, Level2, Level3 and Level4, and then respectively multiplying the characteristics by a weight parameter a4,β4,γ4,λ4And summing to obtain the fused features, wherein the formula represents:
adding the weight parameters to ensure that the characteristics of all layers are the same and the channel number is the same, adjusting the up-sampling or down-sampling channel number of different characteristic layers, obtaining the weight parameters through convolution of a characteristic diagram after resize by 1 × 1, and obtaining the parameter a4,β4,γ4,λ4After concat, limiting the value range to [0, 1 ] by a SoftMax function]And the sum is 1;
6. model evaluation, namely continuously adjusting parameters in a network, optimizing the loss function loss value to the minimum, storing different models according to loss reduction change, and judging through evaluation indexes such as accuracy (Precision), Recall rate (Recall), IoU (interaction Unit), mAP (mean Average precson) and the like;
7. comprehensively judging by the evaluation mode, and selecting a model with the best convergence fitting effect;
8. inputting an image to be detected into a trained model, carrying out model reasoning and identification on the type and position of a specific target of the image and the credibility of the target, judging according to preset alarm setting in the system, and sending early warning information to related power transmission line inspection workers through an alarm system when the specific target with potential harm, such as 'steam hanging', appears, wherein the equipment commonly used in the place, such as a mobile phone APP, is properly processed by the inspection workers according to actual conditions;
9. fig. 3 shows a specific target detected in a certain item, a detection box frame of the target at the lower left corner near the shooting device is large, and a box frame of the targets far away from the shooting device in fig. 4 and 5 is small; the improved Huge-YOLO v3 model based on YOlO v3 can well adapt to the problem of large scale change, and the recognition accuracy effect is obvious.
Claims (2)
1. A power transmission line target detection and monitoring method based on deep learning is characterized by comprising the following steps:
(1) acquiring an image of a specific target through monitoring shooting equipment on a power transmission line site;
(2) screening the images in the step (1) according to a positive sample-negative sample ratio of 3: 7-5: 5, selecting images with proper angles, and deleting images without targets or with unclear angles, wherein the positive samples are images containing specific targets, and the negative samples are images without specific targets;
(3) labeling specific targets in the screened images by using a LabelImg tool, and dividing the images into a training set, a verification set and a test set;
(4) amplifying an image sample before model training, and amplifying the image by adopting a scaling, rotating, cutting or turning mode;
(5) constructing a model on a Darknet deep learning framework by using an improved Huge-YOLO v 3;
(6) training by means of a GPU Nvidia GTX 1080Ti hardware environment based on the constructed model; continuously adjusting parameters in the training process, optimizing the loss function loss value to the minimum, storing different models according to loss reduction change, performing comprehensive judgment through accuracy, recall rate, IoU and mAP evaluation indexes, and selecting the model with the best convergence fitting effect as a detection and monitoring model;
(7) inputting the image to be detected into a trained detection and monitoring model, carrying out model reasoning and identification on the type and position of the specific target of the image and the credibility of the target, sending early warning information to related power transmission line inspection workers through an alarm system when the specific target with potential harm appears, and processing the specific target according to the actual condition by the inspection workers.
2. The deep learning-based power transmission line target detection and monitoring method according to claim 1, characterized in that: the improvement of Huge-Yolo v3 is that the size of an input image is changed from 416 × 416 × 3 to 832 × 832 × 3, a network layer is added on a main structure Darknet-53, and four feature maps of 104 × 104, 52 × 52, 26 × 26 and 13 × 13 are respectively extracted; and fusing the four-layer characteristic diagram by adopting a self-adaptive characteristic fusion mode to obtain fused characteristics, wherein the formula is as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010323443.XA CN111639526A (en) | 2020-04-22 | 2020-04-22 | Power transmission line target detection and monitoring method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010323443.XA CN111639526A (en) | 2020-04-22 | 2020-04-22 | Power transmission line target detection and monitoring method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111639526A true CN111639526A (en) | 2020-09-08 |
Family
ID=72330756
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010323443.XA Pending CN111639526A (en) | 2020-04-22 | 2020-04-22 | Power transmission line target detection and monitoring method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111639526A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112200225A (en) * | 2020-09-23 | 2021-01-08 | 西南交通大学 | Steel rail damage B display image identification method based on deep convolutional neural network |
CN112508030A (en) * | 2020-12-18 | 2021-03-16 | 山西省信息产业技术研究院有限公司 | Tunnel crack detection and measurement method based on double-depth learning model |
CN112528971A (en) * | 2021-02-07 | 2021-03-19 | 北京智芯微电子科技有限公司 | Power transmission line abnormal target detection method and system based on deep learning |
CN112686124A (en) * | 2020-12-25 | 2021-04-20 | 朗坤智慧科技股份有限公司 | Power plant coal conveying belt coal piling detection method and device based on 5G network |
CN112819756A (en) * | 2021-01-15 | 2021-05-18 | 江苏理工学院 | PCB surface defect detection device and method |
CN113255797A (en) * | 2021-06-02 | 2021-08-13 | 通号智慧城市研究设计院有限公司 | Dangerous goods detection method and system based on deep learning model |
CN113611004A (en) * | 2021-08-06 | 2021-11-05 | 寰宇鹏翔航空科技(深圳)有限公司 | Data preprocessing method, device, system and storage medium |
CN113780237A (en) * | 2021-09-27 | 2021-12-10 | 深圳供电局有限公司 | External damage prevention early warning method, device and system for underground pipeline |
CN116596904A (en) * | 2023-04-26 | 2023-08-15 | 国网江苏省电力有限公司泰州供电分公司 | Power transmission detection model construction method and device based on adaptive scale sensing |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390784A (en) * | 2019-07-19 | 2019-10-29 | 国网河北省电力有限公司电力科学研究院 | A kind of transmission line of electricity external force damage prevention monitoring system based on deep learning |
CN110543986A (en) * | 2019-08-27 | 2019-12-06 | 广东电网有限责任公司 | Intelligent monitoring system and monitoring method for external hidden danger of power transmission line |
CN110598757A (en) * | 2019-08-23 | 2019-12-20 | 国网山东省电力公司电力科学研究院 | Detection method for hidden danger of construction machinery of power transmission line |
-
2020
- 2020-04-22 CN CN202010323443.XA patent/CN111639526A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390784A (en) * | 2019-07-19 | 2019-10-29 | 国网河北省电力有限公司电力科学研究院 | A kind of transmission line of electricity external force damage prevention monitoring system based on deep learning |
CN110598757A (en) * | 2019-08-23 | 2019-12-20 | 国网山东省电力公司电力科学研究院 | Detection method for hidden danger of construction machinery of power transmission line |
CN110543986A (en) * | 2019-08-27 | 2019-12-06 | 广东电网有限责任公司 | Intelligent monitoring system and monitoring method for external hidden danger of power transmission line |
Non-Patent Citations (1)
Title |
---|
SONGTAO LIU ET AL.: "Learning Spatial Fusion for Single-Shot Object Detection", 《ARXIV》, pages 1 - 10 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112200225A (en) * | 2020-09-23 | 2021-01-08 | 西南交通大学 | Steel rail damage B display image identification method based on deep convolutional neural network |
CN112508030A (en) * | 2020-12-18 | 2021-03-16 | 山西省信息产业技术研究院有限公司 | Tunnel crack detection and measurement method based on double-depth learning model |
CN112686124A (en) * | 2020-12-25 | 2021-04-20 | 朗坤智慧科技股份有限公司 | Power plant coal conveying belt coal piling detection method and device based on 5G network |
CN112819756B (en) * | 2021-01-15 | 2023-07-11 | 江苏理工学院 | PCB surface defect detection device and method |
CN112819756A (en) * | 2021-01-15 | 2021-05-18 | 江苏理工学院 | PCB surface defect detection device and method |
CN112528971B (en) * | 2021-02-07 | 2021-06-04 | 北京智芯微电子科技有限公司 | Power transmission line abnormal target detection method and system based on deep learning |
CN112528971A (en) * | 2021-02-07 | 2021-03-19 | 北京智芯微电子科技有限公司 | Power transmission line abnormal target detection method and system based on deep learning |
CN113255797A (en) * | 2021-06-02 | 2021-08-13 | 通号智慧城市研究设计院有限公司 | Dangerous goods detection method and system based on deep learning model |
CN113255797B (en) * | 2021-06-02 | 2024-04-05 | 通号智慧城市研究设计院有限公司 | Dangerous goods detection method and system based on deep learning model |
CN113611004A (en) * | 2021-08-06 | 2021-11-05 | 寰宇鹏翔航空科技(深圳)有限公司 | Data preprocessing method, device, system and storage medium |
CN113780237A (en) * | 2021-09-27 | 2021-12-10 | 深圳供电局有限公司 | External damage prevention early warning method, device and system for underground pipeline |
CN116596904A (en) * | 2023-04-26 | 2023-08-15 | 国网江苏省电力有限公司泰州供电分公司 | Power transmission detection model construction method and device based on adaptive scale sensing |
CN116596904B (en) * | 2023-04-26 | 2024-03-26 | 国网江苏省电力有限公司泰州供电分公司 | Power transmission detection model construction method and device based on adaptive scale sensing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111639526A (en) | Power transmission line target detection and monitoring method based on deep learning | |
CN107220618B (en) | Face detection method and device, computer readable storage medium and equipment | |
CN110929774B (en) | Classification method, model training method and device for target objects in image | |
CN109101897A (en) | Object detection method, system and the relevant device of underwater robot | |
CN108830188A (en) | Vehicle checking method based on deep learning | |
CN114399719B (en) | Transformer substation fire video monitoring method | |
CN110659601B (en) | Depth full convolution network remote sensing image dense vehicle detection method based on central point | |
CN110852358A (en) | Vehicle type distinguishing method based on deep learning | |
CN112464766A (en) | Farmland automatic identification method and system | |
CN113642474A (en) | Hazardous area personnel monitoring method based on YOLOV5 | |
CN111524113A (en) | Lifting chain abnormity identification method, system, equipment and medium | |
CN112307984B (en) | Safety helmet detection method and device based on neural network | |
CN114821102A (en) | Intensive citrus quantity detection method, equipment, storage medium and device | |
CN116385958A (en) | Edge intelligent detection method for power grid inspection and monitoring | |
CN115937659A (en) | Mask-RCNN-based multi-target detection method in indoor complex environment | |
CN114462469B (en) | Training method of target detection model, target detection method and related device | |
CN110263836B (en) | Bad driving state identification method based on multi-feature convolutional neural network | |
CN113312999B (en) | High-precision detection method and device for diaphorina citri in natural orchard scene | |
CN115082850A (en) | Template support safety risk identification method based on computer vision | |
CN111724338B (en) | Turntable abnormity identification method, system, electronic equipment and medium | |
Qiu et al. | Underwater sea cucumbers detection based on pruned SSD | |
He et al. | CBAM-YOLOv5: a promising network model for wear particle recognition | |
Suksangaram et al. | The System Operates by Capturing Images of the Wall Surface and Applying Advanced Image Processing Algorithms to Analyze the Visual Data | |
Cai et al. | OCR Service Platform Based on OpenCV | |
CN112967335A (en) | Bubble size monitoring method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |